Functional Magnetic Resonance Imaging: BOLD Response
Magnetic resonance imaging (MRI) has become a widely used diagnostic modality for analysis and examination of anatomical structures. With advancements in the field of science and technology, increasing research regarding further diagnostic enhancement of MRI techniques has been accomplished since the past decade. One such advanced MRI technique is functional magnetic resonance imaging (fMRI). While the conventional MRI shows structural and anatomical details, fMRI also helps in examining physiological body processes (especially metabolic and haemodynamic processes). fMRI helps in establishing neuro-physiological abnormalities underlying various pathologies which could not have been possible with conventional MRI examination. fMRI measures the responses of large populations of neurons, rather than that of single cells. In humans, the best resolution now available with fMRI is around one cortical column, which contains some 105 neurons (Rees, Friston, & Koch, 2000). Currently fMRI has become the most widely used method for brain mapping and studying the neuronal basis of human cognition and other pathologies associated with haemodynamic dysfunctioning (ischemic stroke, tumors etc). The ultimate usefulness and interpretation of fMRI is dependent on the BOLD (the blood-oxygenation level dependent) signal and the relationship of BOLD signal with fMRI and with spiking activity of neurons. These aspects of fMRI would be discussed in detail in this paper.
Conventional MRI examination is scientifically based on behavior of hydrogen atoms or protons in a magnetic field and signals produced on the interaction of these protons with non-ionizing radio frequency signals. fMRI further elaborates on this principle by also utilizing the paramagnetic properties of deoxygenated hemoglobin (Ogawa, Lee, Kay, & Tank, 1990). The principle of BOLD (Blood oxygenation-level dependent) signal is based on the fact that when oxygenated hemoglobin, normally diamagnetic and in low-spin state, gives up its oxygen, the resulting deoxygenated hemoglobin is paramagnetic and in high-spin state (Ogawa, et al). Presence of deoxygenated hemoglobin in blood changes the signals which are emitted from the protons (hydrogen ions) in the water molecules surrounding the deoxygenated hemoglobin. This causes difference between magnetic signals created by oxygenated and deoxygenated blood and are detected by a MRI scanner and manifest as ‘BOLD signal’(Ogawa et al) This effect can be accentuated through the use of gradient-echo techniques in high magnetic fields (? 4 Tesla) (Ogawa et al). Thus paramagnetic deoxygenated hemoglobin in venous blood acts as a naturally occurring contrast agent for fMRI (Ogawa et al). Since BOLD contrast depends on the state of blood oxygenation, physiological body events that change the ratio of oxygenated and deoxygenated hemoglobin in the blood can be detected non-invasively through BOLD signals of fMRI. In 1990, Ogawa et al in their study on rats demonstrated the BOLD signal for the first time, by creating changes in blood oxygen levels in rats by inducing them with anesthetics, by creating insulin induced hypoglycemia, and by making them inhale a gas mixture that altered metabolic demand or blood flow in the brain. BOLD contrast has now become an indispensable part of fMRI studies due to its high sensitivity and wide accessibility.
The relationship between the BOLD fMRI signal and the neuronal activity
Though the use of fMRI is becoming widespread, presently the knowledge regarding the relationship between BOLD signals and underlying neuronal activity is still not sufficient enough. Understanding of the precise relationship between fMRI BOLD signal and neuronal spike activity is important for accurate interpretation of fMRI data. According to Ogawa et al, 1990, there is an increase in BOLD MRI signals during increased neuronal activity. This results from elevated oxygen saturation levels in brain leading to reduction in levels of paramagnetic deoxygenated hemoglobin in arterial and venous blood during neuronal activity. Many studies have been done to examine quantitatively the relationship of the BOLD signal to neuronal activity. Two such studies, independently carried out by Heeger, Huk, Geisler, & Albrecht, (2000) and Rees, Friston, & Koch, (2000), tried to show this relationship by comparing the data obtained by fMRI examination performed in humans, with the electrophysiological data (obtained through recordings of cortically placed electrodes) from the monkeys performing the same task (i.e. observing the moving visual stimuli), as humans. These two studied varied in the areas of the brain in which such relationship was observed. Heeger et al showed linear relationship between BOLD signals and neuronal spike activity in area V1, which is a motion-responsive visual cortical area whereas Rees et al demonstrated this relationship in another motion-responsive, visual cortical area V5.Both the studies were based on the assumption that that the areasV1 and V5 are absolutely homologous to monkey’s motion-responsive visual cortical area MT. Neurons in monkey MT area are direction selective. They increase their firing rates when a stimulus moves in a certain ‘preferred’ direction, while the responses are suppressed when the same stimulus moves in the direction opposite to the ‘preferred’.
The stimulus in both the studies consisted of a field of moving dots as shown in figure 1. Motion coherence (strength) of the stimulus was varied from 0% (each dot moves in a random direction) to 100% (all dots move in the same direction) as shown in figure 1. 50 % motion coherence of the stimulus implied, that 50% of dots were moving in same direction while remaining 50% dots were moving in random directions. The responses of MT neurons to the motion stimulus were recorded across the full range of coherence (varying from 0% to 100% coherence). The neuronal spikes were seen to increase linearly with motion coherence (as shown in figure 1) in both the studies. Both Rees et al and Heegar et al in their studies also showed that fMRI BOLD signals in human visual cortical areas V5 and V1 respectively also showed similar linear response with motion coherence. Since both fMRI response and neuronal activity increased linearly with motion coherence, through mathematical calculations it was proven in both the studies that fMRI BOLD responses are proportional to neuronal activity. This linear relationship between the two is evident from figure 2 which shows that the two set of data (neuronal activity on left ordinate and fMRI measurements on right ordinate) are strikingly similar. Both these studies demonstrated a linear relationship between BOLD fMRI signals and neuronal spike activity, although with different proportionality constants. Proportionality constants were found to be 0.4 and 9 in studies by Heegar et al and Rees et al respectively. Heegar et al in their study found that change in fMRI signal of 1% corresponded to a change in average neuronal firing rate of 0.4spikes/second per neuron where as Rees et al in their study found that change in fMRI signal of 1% corresponded to a change in average neuronal firing rate of 9spikes/second per neuron..
In order to show the relationship between fMRI BOLD signal and neuronal activity,
Logothetis (2002) conducted a study in which he simultaneously measured BOLD fMRI response over a small area around the microelectrode tip (inserted in monkey subcortical area V1) and intracortical electrophysiologic neuronal signals from these same microelectrodes placed in the area V1 in monkeys. Logothetis (2002) found moderate to strong correlation between the neuronal activity measured with microelectrodes and the BOLD fMRI signals.
In short, the results of various studies (Heeger et al , 2000; Logothetis,2002; Rees et al, 2000) comparing the relationship between fMRI BOLD signals and neuronal spike activity, presented here clearly show that the BOLD contrast mechanism directly reflects the neural responses elicited by a stimulus presented for a short duration. It can be also seen from these studies that fMRI BOLD responses and neuronal spike activity show a linear relationship with each other for stimulus presented for short duration.
Figure 1.Average firing rate in monkey MT increases linearly with motion coherence.
Source: Heeger, D.J., Huk, A.C., Geisler, W.S., & Albrecht, D.G. (2000). Spikes versus BOLD: what does neuroimaging tell us about neuronal activity? Nature Neuroscience, 3, 631
Figure 2. Graph showing proportional relationship between fMRI responses in human visual cortical area and average neuronal firing rates in homologous area in monkey
Source: Heeger, D.J., Huk, A.C., Geisler, W.S., & Albrecht, D.G. (2000). Spikes versus BOLD: what does neuroimaging tell us about neuronal activity? Nature Neuroscience, 3, 632.
Relationship between different types of neuronal signals and BOLD response
To understand the relative contribution of several types of neuronal signals to the BOLD
response, Logothetis, 2002 compared two types of nerve signals namely, local field potentials (LFPs) and multi-unit activity (MUA) with their associated fMRI responses recorded simultaneously in monkey visual cortex through micro-electrodes. MUA represents the average spiking activity of small neurons within a sphere of approximately one-third of a millimeter radius with the microelectrode at its centre. This signal is produced as a result of synchronous firings of many neurons. It gets enhanced by summation and thus can be detected at a larger distance. LFPs, on the other hand, predominantly reflect localized synaptic events. Thus LFP represents localized cortical input and localized intracortical nerve processing signals while MUA represent the output signal transmitted to various areas of the brain. Logothelis (2002) in his study observed that only LFP signals and not the MUA ones were significantly correlated with the BOLD response. Thus Logothetis, 2002 reached a conclusion that LFPs were better than the MUAs in predicting the fMRI responses. These findings indicate that the BOLD signal primarily measures the input and processing of neuronal information within a localized region in the cortex and not the output signal transmitted to other brain regions.
Factors affecting the accuracy of the linear relationship
The accuracy of the linear relation-ship between BOLD signal and neuronal activity measured in various studies comparing fMRI responses in human brain with electrophysiological responses in animals has been questioned due to the following reasons:
Many studies conducted to find the relationship between fMRI BOLD signal and neuronal activity used anesthetized animals. Comparing anesthetized animals with awake human beings may not give very accurate results as the anesthetic drug can itself, modulate neuronal activity and reduce evoked responses of the neurons in visual areas (Rees et al, 2000).
fMRI measures the responses of large populations of neurons, rather than that of single cells. Thus, the fMRI signal might reflect not only the firing rates of the local neuronal population but also of the neurons with subthreshold activity. Excitation of these neurons with sub threshold activity would not result in an action potential but would nevertheless deplete blood oxygen. This could interfere with accuracy of results obtained by fMRI (Rees et al).
Neural signals are characterized by considerably higher SNR (signal-to sound ratio) in comparison to the BOLD signals. This implies that the extent of neuronal activation in human-monkey fMRI experiments is often underestimated (Logothetis, 2002)
Reason for the linear relationship between BOLD signals and neuronal activation
Research over the past decade or so has established that BOLD contrast depends not only
on blood oxygenation but also on number of physiological parameters like cerebral blood flow (CBF), cerebral blood volume (CBV), and cerebral mean oxygen utilization rate (CMRO2) (Raichle, 1998). CMRO2 is proportional to neural activity and is four times greater in grey matter as compared to that in white matter (Raichle, 1998). An increase in firing of a neuron causes a proportional increase in local synaptic activity in form or transmission of nerve signal, synthesis, and reuptake of neurotransmitters etc. This leads to a proportional increase in metabolic and oxygen demand. In order to meet this increased metabolic demand, blood flow to the active areas of the brain increases. This increased CBF overcompensates for the increased oxygen requirement, so that an oversupply of oxygenated blood is delivered. Also increase in blood flow and glucose utilization in the brain appears to be much more than oxygen utilization during the period of increased neuronal activity (Raichle, 1998). As a result the amount of deoxygenated hemoglobin is reduced during the periods of neuronal activity. BOLD signal is based on the presence of deoxygenated hemoglobin in blood (Raichle, 1998). Depending on the changes in the amount of deoxygenated hemoglobin, either increases or decreases can occur in the BOLD signal in the normal human brain leading to either positive BOLD or negative BOLD signals respectively. A positive BOLD signal results when increase in cerebral blood flow (CBF) exceeds the increase in cerebral mean oxygen utilization rate (CMRO2), resulting in an elevated oxygen saturation level of venous blood and a concomitant decrease in deoxygenated hemoglobin content (Harel, Lee, Nagoka, Kim, & Kim, 2002). On the other hand it has been hypothesized that if CMRO2 exceeds the cerebral flow, there would be a concomitant increase in the deoxygenated hemoglobin content resulting in a negative BOLD signal (Harel et al).
Consistent with the above hypothesis Harel et al (2002) in their study observed that the
areas of brain which experienced an increase in CBV showed positive BOLD signals, while regions exhibiting the prolonged negative BOLD signal underwent a decrease in CBV.
The increased blood flow, which occurs at time of activity, results in an increased supply of oxygenated blood to the active area of brain resulting in a positive BOLD signal (Harel et al)
Negative bold signal is probably due to decreased cerebral blood flow (Harel et al, 2002).
Negative BOLD signal can be of two types: either transient type or prolonged type. Transient negative BOLD signal presents as an initial dip following the onset of stimulus. This initial dip is probably due to the initial, local increase in oxygen consumption that exceeds the increase in CBF (Hurl et al.). This transient dip is normally followed by a positive BOLD signal. The initial transient dip can be explained by the fact that increase in neuronal activity causes highly localized increases in oxygen consumption, which stimulates a vascular response. This vascular response being delayed by several seconds results in an initial dip. The exact neurophysiological reason behind the prolonged negative BOLD signal has yet not been understood properly. The possible neuro-physiological reason behind this prolonged negative BOLD signal could be the suppression of neuronal activity as suggested by Buckner et al (2000, as cited in Harel et al, 2002). However in the study by Harel et al, prolonged negative BOLD signals were observed due to decreased CBF. Also the association between negative BOLD signal and neuronal activity has not been understood yet. It can be assumed that since positive BOLD signal is associated with increased neuronal activity, negative BOLD signal would be associated with reduced or suppressed neuronal activity. However the assumption that negative BOLD signal is associated with neuronal suppression has not been conclusively proven. The study by Harel et al has shown an increase in neuronal spike activity accompanying the negative BOLD signal. Thus other factors like local re-distribution of blood flow within the vascular network may be involved too in the development of negative BOLD signals. Raichle (1998) hypothesized that development of negative BOLD signal could be related to ‘haemodynamic stealing effect’ due to which the blood is diverted to most active areas of the brain leading to a reduced blood flow in adjacent areas of the brain. In the study by Logothetis (2002), negative BOLD signals were seen exclusively in those cortical regions that were not stimulated by the visual stimulus and thus showed decreased CBF. However the functioning of brain is a complex mechanism and to what extent the above mentioned explanation holds true in different areas of the brain is yet to be discovered. A great deal of research is required in the future in order to understand the exact relationship between fMRI BOLD signal and neuronal activity.
The introduction of advanced imaging techniques (such as fMRI), offers increasingly
detailed information regarding the physiological processes in the brain. Such detailed information, no doubt will immensely help the physician in reaching the correct diagnosis. fMRI will also help the physician in understanding the mechanisms underlying various neurological disorders. Proper understanding of the relationship between fMRI signals and neuronal activity is of vital importance for correct interpretation of fMRI data and establishing its relationship with underlying neuronal pathology. Though the studies over the last decade have established a linear relationship between BOLD responses and neuronal responses in relation to stimulus presented for a short duration of time, the information regarding the exact relationship between positive and negative fMRI BOLD response in different areas of the brain and neuronal activity, at present is still conflicting and incomplete. Nevertheless it does provide a basis for initiation of larger studies in future in order to discover the exact relationship between fMRI BOLD response and neuronal activation.
Harel, N., Lee, S.P., Nagoka, T., Kim, D-S., & Kim, S.G. (2002). Origin of negative blood
oxygenation level-dependent fMRI signals. Journal of Cerebral Blood Flow & Metabolism, 22, 908–917.
Heeger, D.J., Huk, A.C., Geisler, W.S., & Albrecht, D.G. (2000). Spikes versus BOLD: what
does neuroimaging tell us about neuronal activity? Nature Neuroscience, 3, 631 – 633.
Logothetis, N.K. (2002). The neural basis of the blood-oxygen-level-dependent functional
magnetic resonance imaging signal. Philosophical Transactions of the Royal Society B: Biological Sciences, 357(1424):1003-1037.
Ogawa, S., Lee, T.M., Kay, A.R, & Tank, D.W. (1990). Brain magnetic resonance imaging
with contrast dependent on blood oxygenation. Proceedings of National Academy of Science of U. S. A , 87, 9868?9872.
Raichle, M.E. (1998). Behind the scenes of functional brain imaging: a historical and
physiological perspective. Proceedings of National Academy of Science of the U S A, 95, 765-772.
Rees, G., Friston, K., & Koch, C. (2000).A direct quantitative relationship between the
functional properties of human and macaque V5. Nature Neuroscience, 3, 716 – 723.